Python list comprehension for three-dimensional array - python

I have a three dimensional array in this format:
x = [
[[1,2,3,4,5],[6,7,8,9,10]],
[[11,12,13,14,15],[16,17,18,19,20]],
[[21,22,23,24,25],[26,27,28,29,30]],
[[21,22,23,24,25]]
]
I'd like to split it into two, three dimensional arrays in this format:
y = [
[[1,2,3],[6,7,8]],
[[11,12,13],[16,17,18]],
[[21,22,23],[26,27,28]],
[[21,22,23]]
]
z = [
[[4,5],[9,10]],
[[14,15],[19,20]],
[[24,25],[29,30]],
[[24,25]]
]
I came up with this list comprehension for creating y:
[j[:3] for i in x for j in i]
Which returns this:
[[1, 2, 3], [6, 7, 8], [11, 12, 13], [16, 17, 18], [21, 22, 23], [26, 27, 28], [31, 32, 33]]
But as you'll see, it doesn't maintain the same multi-dimensional shape. Does anyone have any ideas?

You need to iterate one level deeper:
x = [[[1, 2, 3, 4, 5], [6, 7, 8, 9, 10]], [[11, 12, 13, 14, 15], [16, 17, 18, 19, 20]], [[21, 22, 23, 24, 25], [26, 27, 28, 29, 30]], [[21, 22, 23, 24, 25]]]
y = [[i[:3] for i in b] for b in x]
z = [[i[-2:] for i in b] for b in x]
Output:
[[[1, 2, 3], [6, 7, 8]], [[11, 12, 13], [16, 17, 18]], [[21, 22, 23], [26, 27, 28]], [[21, 22, 23]]]
[[[4, 5], [9, 10]], [[14, 15], [19, 20]], [[24, 25], [29, 30]], [[24, 25]]]

Move your inner most loop into a nested comprehension so that the inner lists are preserved:
y = [[j[:3] for j in i] for i in x]

Related

Numpy. How to split 2D array to multiple arrays by grid?

I have a numpy 2D-array:
c = np.arange(36).reshape(6, 6)
[[ 0, 1, 2, 3, 4, 5],
[ 6, 7, 8, 9, 10, 11],
[12, 13, 14, 15, 16, 17],
[18, 19, 20, 21, 22, 23],
[24, 25, 26, 27, 28, 29],
[30, 31, 32, 33, 34, 35]]
I want to split it to multiple 2D-arrays by grid 3x3. (It's like a split big image to 9 small images by grid 3x3):
[[ 0, 1,| 2, 3,| 4, 5],
[ 6, 7,| 8, 9,| 10, 11],
---------+--------+---------
[12, 13,| 14, 15,| 16, 17],
[18, 19,| 20, 21,| 22, 23],
---------+--------+---------
[24, 25,| 26, 27,| 28, 29],
[30, 31,| 32, 33,| 34, 35]]
At final i need array with 9 2D-arrays. Like this:
[[[0, 1], [6, 7]],
[[2, 3], [8, 9]],
[[4, 5], [10, 11]],
[[12, 13], [18, 19]],
[[14, 15], [20, 21]],
[[16, 17], [22, 23]],
[[24, 25], [30, 31]],
[[26, 27], [32, 33]],
[[28, 29], [34, 35]]]
It's just a sample what i need. I want to know how to make small 2D arrays from big 2D array by grid (N,M)
You can use something like:
from numpy.lib.stride_tricks import sliding_window_view
out = np.vstack(sliding_window_view(c, (2, 2))[::2, ::2])
Output:
>>> out.tolist()
[[[0, 1], [6, 7]],
[[2, 3], [8, 9]],
[[4, 5], [10, 11]],
[[12, 13], [18, 19]],
[[14, 15], [20, 21]],
[[16, 17], [22, 23]],
[[24, 25], [30, 31]],
[[26, 27], [32, 33]],
[[28, 29], [34, 35]]]

Creating 2D list from 3D list in Python [duplicate]

This question already has answers here:
How do I make a flat list out of a list of lists?
(34 answers)
Closed 11 months ago.
Assume I have a 3D list as following:
[[[1, 2, 3], [4, 5, 6], [7, 8, 9]], [[10, 11, 12], [13, 14, 15], [16, 17, 18]], [[19, 20, 21], [22, 22, 23], [24, 25, 26]]]
I want to convert it to:
[[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12], [13, 14, 15], [16, 17, 18], [19, 20, 21], [22, 22, 23], [24, 25, 26]]
I do not want to get lost inside of the for loops, is there a way this can be easily implemented?
Thanks in advance.
Try summing the list with []:
sum(your_3d_list, [])
Attempt it Online!
You can use numpy module and reshape function:
import numpy as np
myList = [[[1, 2, 3], [4, 5, 6], [7, 8, 9]], [[10, 11, 12], [13, 14, 15], [16, 17, 18]], [[19, 20, 21], [22, 22, 23], [24, 25, 26]]]
array = np.array(myList)
array.reshape(9,3)
Output
array([[ 1, 2, 3],
[ 4, 5, 6],
[ 7, 8, 9],
[10, 11, 12],
[13, 14, 15],
[16, 17, 18],
[19, 20, 21],
[22, 22, 23],
[24, 25, 26]])
You can use assert in order to make sure that this is the expected array:
expected = [[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12], [13, 14, 15], [16, 17, 18], [19, 20, 21], [22, 22, 23], [24, 25, 26]]
array = np.array(myList)
assert array.reshape(9,3).tolist() == expected
which works fine!
Note that, array.reshape(9,3) returns a numpy array and not a list. If you want to have the expected array as a list, you can use array.reshape(9,3).tolist()
array.reshape(9,3).tolist()
Output
[[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12], [13, 14, 15], [16, 17, 18], [19, 20, 21], [22, 22, 23], [24, 25, 26]]
Possible solution is the following:
lst = [[[1, 2, 3], [4, 5, 6], [7, 8, 9]], [[10, 11, 12], [13, 14, 15], [16, 17, 18]], [[19, 20, 21], [22, 22, 23], [24, 25, 26]]]
result = [item for subitem in lst for item in subitem]
print(result)
Prints
[[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12], [13, 14, 15], [16, 17, 18], [19, 20, 21], [22, 22, 23], [24, 25, 26]]
​
You can use
import itertools
library, which is a standard python library.
use it like this:
import itertools
array = itertools.chain.from_iterable(array)
array = list(array)
the output will be:
[[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12], [13, 14, 15], [16, 17, 18], [19, 20, 21], [22, 22, 23], [24, 25, 26]]
You can loop through each 2d list in the 3d list, then loop through each 1d list in the 2d list
_3d_list = [[[1, 2, 3], [3, 2, 1], [1, 3, 5]], [[4, 5, 6], [6, 5, 4], [4, 6, 8]]]
final_list = []
for _2d_list in _3d_list:
for _1d_list in _2d_list:
final_list.append(_1d_list)
print(final_list)
Output:
[[1, 2, 3], [3, 2, 1], [1, 3, 5], [4, 5, 6], [6, 5, 4], [4, 6, 8]]

what is the difference between two following reshape function in numpy?

I am building a neural network. where I have to flatten my training dataset.
I have two options.
1 is:
train_x_flatten = train_x_orig.reshape(train_x_orig.shape[0], -1).T
and 2nd one is:
train_x_flatten = train_x_orig.reshape(train_x_orig.shape[1]*train_x_orig.shape[2]*train_x_orig.shape[3], 209)
both gave the same shape but I found difference while computing cost?
why is that? thank you
Your original tensor is of at least rank 4 based on the second example. The first example pulls each element, ordered by increasing the right-most index, and inserts the elements into rows the length of the zeroth shape. Then transposes.
The second example again pull elements from by incrementing from the right-most index, i.e.:
element = train_x_orig[0, 0, 0, 0]
new_row.append(element)
element = train_x_orig[0, 0, 0, 1]
new_row.append(element)
but the size of the row is different. It is now the dimension of everything else in the tensor.
Here is an example to illustrate.
First we create an ordered array and reshape it to rank 4.
import numpy as np
x = np.arange(36).reshape(3,2,3,2)
x
# returns:
array([[[[ 0, 1],
[ 2, 3],
[ 4, 5]],
[[ 6, 7],
[ 8, 9],
[10, 11]]],
[[[12, 13],
[14, 15],
[16, 17]],
[[18, 19],
[20, 21],
[22, 23]]],
[[[24, 25],
[26, 27],
[28, 29]],
[[30, 31],
[32, 33],
[34, 35]]]])
Here is the output of the first example
x.reshape(x.shape[0], -1).T
# returns:
array([[ 0, 12, 24],
[ 1, 13, 25],
[ 2, 14, 26],
[ 3, 15, 27],
[ 4, 16, 28],
[ 5, 17, 29],
[ 6, 18, 30],
[ 7, 19, 31],
[ 8, 20, 32],
[ 9, 21, 33],
[10, 22, 34],
[11, 23, 35]])
And here is the second example
x.reshape(x.shape[1]*x.shape[2]*x.shape[3], -1)
# returns:
array([[ 0, 1, 2],
[ 3, 4, 5],
[ 6, 7, 8],
[ 9, 10, 11],
[12, 13, 14],
[15, 16, 17],
[18, 19, 20],
[21, 22, 23],
[24, 25, 26],
[27, 28, 29],
[30, 31, 32],
[33, 34, 35]])
How the elements get reordered is fundamentally different.

How to sum elements inside a nested list?

So, I have a list that looks something like this
[[[1, 2], [3, 4], [5, 6]],
[[7, 8], [9, 10], [11, 12]],
[[13, 14], [15, 16], [17, 18]],
[[19, 20], [21, 22], [23, 24]]]
And I want it to look like this
[[3, 7, 11],
[15, 19, 23],
[27, 31, 35],
[39, 43, 27]]
that is 3 = sum([1, 2]), 7 = sum([3, 4]), ....
I have tried nesting for loops but I haven't found anything that got the desired result, does anyone know how I could do this?
This will do the job quite nicely and imo is more readable than list comprehensions.
lists = [[[1, 2], [3, 4], [5, 6]],
[[7, 8], [9, 10], [11, 12]],
[[13, 14], [15, 16], [17, 18]],
[[19, 20], [21, 22], [23, 24]]]
new_lists = []
for nested in lists:
new_ls = []
for ls in nested:
new_ls.append(sum(ls))
new_lists.append(new_ls)
>>> new_lists
[[3, 7, 11], [15, 19, 23], [27, 31, 35], [39, 43, 47]]
You could also use list comprehensions:
[[sum(x) for x in triple] for triple in lists]
In the above list comprehension, triple will be your list of three doubles so the first for loop will be covering these. x will then be each list of doubles, inside of the triple so we are summing it, while keeping it inside the original triple by using this bracket around:
[sum(x) for x in triple]
output:
[[3, 7, 11], [15, 19, 23], [27, 31, 35], [39, 43, 47]]
If you are happy to use a 3rd party library, you can use NumPy and sum along a single dimension:
import numpy as np
A = np.array([[[1, 2], [3, 4], [5, 6]],
[[7, 8], [9, 10], [11, 12]],
[[13, 14], [15, 16], [17, 18]],
[[19, 20], [21, 22], [23, 24]]])
res = A.sum(2)
Result:
array([[ 3, 7, 11],
[15, 19, 23],
[27, 31, 35],
[39, 43, 47]])
See also: What are the advantages of NumPy over regular Python lists?

Move elements out of sublists into new list

I have a list that holds several sublist, each one of them with a given number of elements inside. I need to move all elements inside all sublists into another list, ie: remove the separation among elements imposed by the sublists.
This is a MWE if what I mean:
a = [[[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], [13, 14, 15, 16]], [[17, 18, 19, 20], [21, 22, 23, 24]], [[25, 26, 27, 28], [26, 30, 31, 32], [33, 34, 35, 36]]]
b = []
for elem in a:
for item in elem:
b.append(item)
which results in:
[[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], [13, 14, 15, 16], [17, 18, 19, 20], [21, 22, 23, 24], [25, 26, 27, 28], [26, 30, 31, 32], [33, 34, 35, 36]]
I'm sure there's a more elegant and simpler way to do this in python.
Use itertools.chain.from_iterable:
>>> from itertools import chain
>>> list(chain.from_iterable(a))
[[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], [13, 14, 15, 16], [17, 18, 19, 20], [21, 22, 23, 24], [25, 26, 27, 28], [26, 30, 31, 32], [33, 34, 35, 36]]
Timing comparison:
Try this:
[item for sublist in a for item in sublist]

Categories

Resources